One of the perennial debates in racing is over the effects of the barrier draw on a horse’s chances in a race.

In August 2003, I joined in a discussion on the web-based racing forum, AusRace, regarding barriers.

Peter Harrop, a regular contributor, wrote: “I think the only correct way to quote barrier statistics is to put in place a minimum field size. If you only include races with field sizes of 12 horses or more, then the barrier statistics of barriers one to 12 have relevance and are comparable. When this is done on NZ races, there is a slight correlation between barrier draw and winningness, but the decline on the graph isn’t enough to make a spring slinky work.


“I believe barrier draws are a totally over-rated concept. In the ratings I do, strike rates only worsen when I try to factor them in. Essentially though, if anything, I think the very outside draw of any field is as good as any.”

PPM stalwart, E.J. Minnis replied: “... Using your guidelines re races with 12 or more runners, the inside barriers still win more than the outside barriers. Maybe the skew is not as great as some would believe, but a skew does exist.

An analysis of 26,500 races with a minimum of 12 and a maximum of 16 runners discloses that the inside 50 per cent of barriers win 52.5 per cent of races (in races with an odd number of runners of either 13 or 15 runners, the middle barrier horse has been disregarded in the analysis).

“However, as the number of runners increases, so does the skew towards the inside 50 per cent of barriers, i.e. in 12 runner races, 52.5 per cent of races are won by horses drawn on the inside, while in 16 runner races, 55.6 per cent of races are won by horses drawn on the inside.”

The stats put up in this discussion would be familiar to keen form students. What’s more, they hold up as impact ratios comparing individual strike rates of inside vs outside barriers. With that kind of evidence for inside barriers why would anyone bother querying the barrier issue further?

My answer is because 300,000 or so horses from 26,500 races still don’t explain “why” the strike rates pan out that way, or if there are any effects of human doings in the figures.

The figures showed clearly the inside-50 per cent of barriers had a higher strike rate than the outside-50 per cent.

Why might outer barrier horses perform more poorly than inside barrier ones? I had a theory the strike rate difference had to do with horse quality more than inside or outside draw. I believed the number of BELOW average quality horses that drew the outside-50 per cent of barriers was greater than the number of ABOVE average quality horses that drew the outside-50 per cent of barriers.

In my investigation of this, I began with below and above average quality of horse being defined by average handicap weight in a race. In what follows, BARHWt means Below Average Race Handicap Weight, and AARHWt means Above Average Race Handicap Weight.

I argued at the time, there was a higher than 50 per cent chance BARHWt horses would fill most of the outside-50 per cent of barriers. I based that on a near certain sense that BELOW average quality horses outnumber ABOVE average horses in races overall. It was quickly pointed out that on a random draw basis there was exactly the same chance of BARHWt horses filling inside barriers as outside ones.

This was true of course. But it wouldn’t disprove what I was trying to show, which was that more BARHWt horses drew the outside-50 per cent of barriers than AARHWt horses. I thought BARHWt horses probably UNDER-PERFORMED from wherever they drew while AARHWt horses OVER-PERFORMED from all their barrier draws. Furthermore, that the effects wouldn’t be seen until bigger fields (13+).

To satisfy my curiosity about likely numbers of BARHWt and AARHWt horses drawing the inside and outside-50 per cent, on Saturday, August 9, I conducted a preliminary investigation. The result was what I had hoped. Using just the 15 Saturday handicap races around the country that had 13 or more starters that day, I counted 219 runners of which 79 were AARHWt horses and 140 were BARHWt. Of the 79 AARHWts, 40 drew inside-50 per cent and 37 drew outside-50 per cent.

Of the 140 BARHWts, 72 drew inside-50 per cent and 63 drew outside-50 per cent (mid barrier horses were not counted). The outside BARHWts (63) did not outnumber the inside ones (72) but the 63 outside BARHWts did significantly outnumber the 37 outside AARHWts. Things were looking good for my hunches.

Crunched database numbers don’t show up the quality of horse. In crunched numbers outside-50 per cent BARHWt horses and inside-50 per cent BARHWt horses are just simple amounts.

The data needed to answer my  questions about the beliefs outlined had to include horse quality. Since as many BARHWt horses randomly draw INSIDE as OUTSIDE, the task wasn’t about proving the 52.5 per cent to 55.6 per cent inside strike rate was untrue, or true, but to account for why it was.

The original conclusion from the 26,500 races of 12-16 runners inclusive, implied that the strike rate for inner-drawn horses was because of inside draws. The figures mentioned no other variable associated with the draw. Again, I wondered if trainer behaviour played a part.

So, over a period of nine months from August 2003 I accumulated statistics as races unfolded rather than by retro-checking data, to see if lower quality runners did predominate in races and how they performed from inside compared to outside barriers.

To test my theory:

  1. I defined horse “quality” within an individual race as determined by handicap weight. (Note: Quality could just as easily have been defined by number of horses on limit weight versus those above limit weight, per race, by $API, by career WSR per cent, etc. There are several ways to set up an operational definition of individual runner quality.)
  2. I averaged the handicap weight (not the carried weight) in each handicap race of Saturday and public holiday meetings (using metro venues SR MR PR BR AR) and separated runners in each race into those that were AARHWt and those that were BARHWt. Occasionally a horse’s handicap weight equalled exactly the race average handicap weight...these horses formed a separate group, were counted, but not as either above or below average.
  3. I tagged horses ABOVE average race handicap weight AARHWt; Those BELOW race average handicap weight, BARHWt. All set weight and WFA races (including Maidens where set weights apply) were excluded from the sampling as weights in these races are not allocated on merit or quality.
  4. I then compared winner category (AARHWt or BARHWt) with their barrier draw, splitting inside-50 per cent barriers from outside-50 per cent of barriers.
  5. I looked only at bigger fields of 13 and upwards, believing the effect of wide draw would be more exaggerated in bigger fields, and may enable direct comparison to see if there was a skew as field size increased.
  6. I considered counting scratchings due to trainer dissatisfaction with barrier but had no “hard” way to establish motive.

So, with data-gathering para­meters set to these conditions, this is what I collected from 159 races, covering 2,358 runners with an average of 14.83 runners per race:

AARHWt winning horses drawn inside-50 per cent 43
AARHWt winning horses drawn outside-50 per cent 49
BARHWt winning horses drawn inside-50 per cent 32
BARHWt winning horses drawn outside-50 per cent 31
Either type that were mid barrier (in fields of 13, 15, etc.) and won 4
Number of horses sampled = 2,358
Number of races sampled = 159
Average field size 14.83
Per cent of horses of 2,358 that were AARHWt = 37.3 per cent (879)
Per cent of horses of 2,358 that were BARHWt = 59.4 per cent (1,401)
Per cent of horses that were exact RHWt = 3.3 per cent (78)

ANALYSIS/DISCUSSION
It may be argued that classifying horses as above or below average by using the average race handicap weight is not a true measure of above average quality or below average quality. I concede this. However, even if “quality” were decided by comparisons of career WSR per cent or $APIs, it would not materially alter the findings in my sample because there is a strong correlation between handicap weight and these other measures of “quality” anyway.

  • BARHWt made up the MAJORITY of runners. That is, 59.4 per cent of 2,358 or 14,01 runners, yet they produced only 63 winners (39.62 per cent).
  • AARHWt horses were in the MINORITY – 879 runners (37.4 per cent), yet they produced 92 winners (57.86 per cent of all winners) from that number of runners. (On a proportional basis this is equivalent to 146.6 AARHWt winners to 63 BARHWt winners aeq. at 1,401 runners apiece.) This is a significant difference between AARHWt and BARHWt runners! It implies strongly that class/ability overrides barrier draw.
  • The outside-50 per cent horses (of either type, AARHWt and BARHWt) collectively, produced 80 of the 159 wins or 50.31 per cent of all winners. The inside-50 per cent of horses (of either type, AARHWt and BARHWt) produced 75 of the 159 winners or 47.16 per cent of all winners. These figures showing outside-50 per cent barriers striking at a higher rate than inside-50 per cent were something of a surprise as the strike rate for outside-50 per cent in the 26,500 race sample was 47.5 per cent and lower.
  • The most telling statistic of all however, was the win rate of BARHWt horses. 32 of 159 inside-50 per cent BARHWt horses won (20.1 per cent of all races); 31 of 159 of outside-50 per cent BARHWt horses won (19.5 per cent of all races). If anything, this shows CLEARLY that lower ability horses (BARHWt) perform more poorly than higher ability horses wherever drawn.
  • If we look at just the outside-50 per cent BARHWt horses, the result suggests the bigger the fields are, the more likely the strike rate of these horses is likely to remain low (19.5 per cent WSR 31/159) esp­ecially compared to wider-drawn AARHWt horses (30.8 per cent WSR 49/159). So, these figures show conclusively BARHWt horses out number AARHWt horses generally (1,401 vs 879) and that BARHWt horses that jump from outside-50 per cent barriers outnumber AARHWt horses from outside-50 per cent barriers.

Is this enough to explain why wide barriers have a lower strike rate than inside ones? After all, many punters and especially trainers hold that horses perform better from inside draws. Some even accept inside barriers “cause” horses to go better. The 52.5 per cent to 55.6 per cent inside strike rates of the 26,500 race sample seem to support some kind of inside favouring. My figures above may go some way towards explaining “why”, but what of the quirky trainer factor?

Perhaps the definitive reason for the poorer strike rate for outside-50 per cent horses has to do with human intervention PLUS the number of AARHWt horses that jump from the inside-50 per cent. Consider this – inside draws of themselves, do NOT favour below average horses at all in my BARHWt statistics. Below average quality horses performed with about the same strike rate WHEREVER THEY DREW, and always lower than above average quality horses...which performed fairly well from everywhere.

Why do I mention the AARHWts drawn inside-50 per cent? Not showing clearly in my figures is the fact that slightly more AARHWt horses draw inside-50 per cent compared to outside-50 per cent. Why is this? Anecdotally, it’s well-known trainers scratch horses with unfavourable draws and renominate or double nominate to get better barriers (usually inside ones). This practice is in widespread use in Australia and possibly explains the slight skew of AARHWt horses to inside barriers. It shows up clearly in stats of UK racing areas that allow preferential barrier draws. It suggests there is more than a random 50-50 chance of AARHWt horses drawing inside-50 per cent barriers.

Trainer manipulated barrier placement of AARHWt horses may put an extra +1 per cent to +2 per cent on the win rate for inside barriers, robbing the same percentage from outside barriers. This may not seem much but effectively takes inside v outside win stats from 50 per cent-50 per cent to 52 per cent – 48 per cent. And that may be the needed last piece of the puzzle.

The large number of BARHWt runners drawn wide, plus above average quality horses massaged into preferred inside draws by workings of the nomination system by trainers, possibly accounts for the higher strike rate of inside barriers. Feel free to disagree.

Hoofnote:
In this exercise I excluded from the count quite a number of “straight course” races at Flemington where barrier draw is arguably irrelevant, as well as all set weight races whether WFA, Special Conditions or Maiden class. Four races in the sample of 159 races had winners which were the exact RHWt and drew the exact mid barrier of an odd-numbered field (13, 15).

These four races were included in the 159 races of the sample, but the winner was not credited as being either outside-50 per cent or inside-50 per cent. In addition to this, I excluded from the final sample of 159, a number of races where winners, mainly AARHWt and mainly outside-50 per cent barriers, that had weight that was less than 0.5kg above/below the race average handicap weight. This sample of 159 races and 2,358 runners is as squeaky clean as I could make it from the control angle.

By Tony Acciano

PRACTIAL PUNTING - OCTOBER 2004